I made some progress for the terrain generation to get rid of the checkboard pattern.
My first attemp was the globcover image I used in the Minecraft Earth Map (external link), which has a resolution of 300 m. And here is the biggest problem: every pixel represents more than 1 block ingame, resulting in a chessboard pattern like this.
You can see that streets and rivers are smooth, because the osm data is vector based and not pixel based, but the terrain looks horrible.
NEW TERRAIN GENERATION
Now the script uses a grayascale vegetation map, so we can apply an interpolation and blur effekt, to fill the pixel in betweens for higher scales (most global coverage maps are way to small for this project). In Addition, I use ImageMagick to convert the grayscal image into a monochrom image (with dithering). This way, we get an image with only the colours black and white. In WorldPainter, we match “black” with sand and “white” with grass.
Vegetation map for underground terrain distribution:
And here you can see an example on how dithering works:
(From grayscale image to monochrom image using different technics)
In the next step, I use the “Earth without clouds, Arctic Ocean ice, or shaded relief” from ShadedRelief (external link). I filtered out some images to get areas where the ground coverage has a very red-ish colour and a very grey-ish colour. Again we use interpolation and dithering afterwards to increase the resolution. In the WorldPainter script, these masks are used to replace sand with red sand or gravel (grass will stay the same).
Later in the script, we can use OSM data for places like urban areas, farms, and meadows. In Addition, on slopes with a certain degree, the script applies stone as a ground material. I’m planning to change this a bit. This way grass, sand, red sand and gravel will have other corresponding blocks on steep terrain.
Here you can see an older images of the Neil Delta. The saturation of the gravel is way to high in my first attempt and it is already lowered.
Furthermore you can see a lot of roads, rivers of different size and a lot of different landuses, for example farms (at least all the farms that are available in the OSM data).
Last but not least, an interactive map of Silicy and the southern part of Italy: